TBShoNet provides a method to develop an algorithm that can be deployed on phones to assist healthcare providers in areas where radiologists and high-resolution digital images are unavailable
A study titled, Transfer Deep Learning for Tuberculosis Detection on Chest X-Ray Images Captured by Phone Camera, presented at the ongoing Radiological Society of North America (RSNA) 106th Scientific Assembly and Annual Meeting found that a deep learning-based tuberculosis (TB) detection model can detect TB on phone-captured chest X-ray photographs.
An early diagnosis of TB is crucial but challenging for resource-poor countries. TBShoNet provides a method to develop an algorithm that can be deployed on phones to assist healthcare providers in areas where radiologists and high-resolution digital images are unavailable. This is the first study applying transfer deep learning to smartphone-captured chest X-ray photos for TB diagnosis.
The study authors Po-Chih Kuo, Ph.D., presenter, Cheng-Che Tsai,M.D., Leo Anthony Celi, M.D. used three publicly available datasets for model pre-training, transferring and evaluation. The neural network was pre-trained on a database containing 250,044 chest X-rays with 14 pulmonary labels, which did not include TB.
The model was then recalibrated for chest X-ray photographs by using simulation methods to augment the dataset. The TBShoNet model was built by connecting the pre-trained model to an additional 2-layer neural network trained on augmented chest X-ray images. Then 662 chest X-ray photographs taken by five different phones (TB: 336; normal:326) were used to test the model performance. Sensitivity and specificity for TB classification were 81 per cent and 84 per cent, respectively.